CAD Plugins

Agent-ready CAD skills and Python utilities for DWG/DXF conversion, quick summaries, focused inspection, and readable previews.

한국어 | English

Quickstart | Skills | Commands | Examples | Workflow

---

Why This Exists

CAD files are rarely just pictures. A single drawing can contain nested blocks, layers, dimensions, labels, attributes, construction geometry, and huge coordinate spaces. CAD Plugins gives agents a practical CAD workflow:

• Preserve original DWG/DXF inputs.
• Normalize DWG to canonical DXF when needed.
• Parse CAD structure into annotation-aware JSON.
• Render purpose-specific previews instead of one overloaded screenshot.
• Inspect by layer, block, text, entity type, keyword, or coordinate region.

Skills

Area	Path	Purpose
Conversion	skills/cad-convert	Normalize DWG/DXF inputs into canonical DXF and validate results.
Quicklook	skills/cad-quicklook	Produce concise first-pass summaries for non-CAD specialists.
Rendering	skills/cad-render	Generate readable PNG/SVG/HTML previews from CAD analysis JSON.
Inspection	skills/cad-inspect	Extract focused answers about layers, labels, blocks, entities, and regions.
Python utilities	cad_parser/	Shared conversion, parsing, modeling, and visualization code used by the skills.

Quickstart

git clone https://github.com/GirinMan/cad-plugins.git
cd cad-plugins

python3 --version  # must be Python 3.10+
python3 -m venv .venv
source .venv/bin/activate

python -m pip install -U pip
python -m pip install -r requirements.txt
python -m unittest discover -v

If macOS python3 points to Python 3.9, use a newer interpreter such as python3.13 -m venv .venv.

Windows PowerShell:

git clone https://github.com/GirinMan/cad-plugins.git
cd cad-plugins

py -3.11 -m venv .venv
.\.venv\Scripts\Activate.ps1

python -m pip install -U pip
python -m pip install -r requirements.txt
python -m unittest discover -v

Any Python 3.10+ interpreter is fine on Windows; py -3.11 is just a concrete example.

Install As Skills

Codex local skill install:

mkdir -p ~/.codex/skills
cp -R skills/cad-* ~/.codex/skills/

Claude-compatible local skill copy:

mkdir -p ~/.claude/skills
cp -R skills/cad-* ~/.claude/skills/

Windows PowerShell:

New-Item -ItemType Directory -Force "$env:USERPROFILE\.codex\skills"
Copy-Item -Recurse -Force .\skills\cad-* "$env:USERPROFILE\.codex\skills\"

New-Item -ItemType Directory -Force "$env:USERPROFILE\.claude\skills"
Copy-Item -Recurse -Force .\skills\cad-* "$env:USERPROFILE\.claude\skills\"

The plugin manifest lives at .codex-plugin/plugin.json and points to ./skills/.

CAD Converter Dependencies

DXF files can be parsed directly with Python. DWG conversion needs a local converter.

Recommended DWG converter:

• ODA File Converter for macOS and Windows.
• LibreDWG dwgread as an open-source fallback when available, with per-file validation.

macOS ODA path example:

export CAD_PARSER_ODA_PATH="/Applications/ODAFileConverter.app/Contents/MacOS/ODAFileConverter"
python -m cad_parser.convert --check

Windows ODA path example:

$env:CAD_PARSER_ODA_PATH = "C:\Program Files\ODA\ODAFileConverter\ODAFileConverter.exe"
python -m cad_parser.convert --check

If you need the Windows installer itself, use the ODA File Converter MSI and install it first:

https://www.opendesign.com/guestfiles/get?filename=ODAFileConverter_QT6_vc16_amd64dll_27.1.msi Persist it for future Windows shells if desired:

[Environment]::SetEnvironmentVariable(
  "CAD_PARSER_ODA_PATH",
  "C:\Program Files\ODA\ODAFileConverter\ODAFileConverter.exe",
  "User"
)

Common Commands

Check converter availability:

python -m cad_parser.convert --check

Convert one DWG to canonical DXF:

python -m cad_parser.convert input.dwg \
  --out data/intermediate/input.dxf \
  --manifest data/processed/input_conversion_manifest.json

Analyze DXF or DWG into JSON and PNG:

python -m cad_parser.cli input.dxf \
  --out data/processed/input_analysis.json \
  --png-out data/processed/input_preview.png \
  --png-width 2400 \
  --png-height 1800 \
  --png-clip-percentile 0.01

Render again from an existing JSON analysis:

python -m cad_parser.visualize data/processed/input_analysis.json \
  --png-out data/processed/input_linework.png \
  --width 2400 \
  --height 1800 \
  --clip-percentile 0.01

Examples

B010 수치지도 DXF

This sample behaves like a map-style DXF: layer-centered geometry with meaningful labels. A quicklook/render pass produces a readable clipped preview suitable for orientation before deeper label inspection.

Pump Station Plan DWG

This DWG is normalized to DXF first, then rendered as CAD-derived previews. The useful view is not a single full screenshot: clipped linework and block-aware previews make the dense plan easier to inspect.

Block-aware preview:

Recommended Agent Workflow

1. Start with cad-convert for every DWG/DXF input so downstream steps use canonical DXF.
2. Use cad-quicklook when the user asks "what is this file?" or needs a short summary.
3. Use cad-render when the user needs to see the drawing, but pick a purpose-specific view.
4. Use cad-inspect only after the target is specific: keyword, layer, block, coordinate area, or facility label.

What Not To Commit

Do not commit customer CAD originals, huge normalized JSON reports, or generated intermediate DXF files by default. Put reusable public examples under assets/examples/ after confirming they are safe to share.

Verification

Before publishing changes, run:

python -m json.tool .codex-plugin/plugin.json >/dev/null
python -m unittest discover -v
python ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/cad-convert
python ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/cad-quicklook
python ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/cad-render
python ~/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/cad-inspect